Académique Documents
Professionnel Documents
Culture Documents
ABSTRACT:A moment inequality is derived for the system whose life distribution is in an overall decreasing life
(ODL) class of life distributions. A new nonparametric test statistic for testing exponentiality against ODL is
investigated based on this inequality. The asymptotic normality of the proposed statistic is presented. Pitman's
asymptotic efficiency, power and critical values of this test are calculated to assess the performance of the test.
Real examples are given to elucidate the use of the proposed test statistic in the reliability analysis. Wealso
proposed a test for testing exponentiality versus ODL for right censored data and the power estimates of this test
are also simulated for censored data for some commonly used distributions in reliability. Finally, real data are
used as an example for practical problems.
1
= 0 .
Sepehrifar et al. [30] defined the ODL .And investigated the probabilistic characteristics of this class of life
distribution.
www.ijmsi.org 62 | Page
A Moment Inequality For Overall Decreasing Life Class Of Life Distributions With Hypotheses
Definition 1.1: A life distribution Fon 0, , with 0 = 0 is called overall decreasing life (ODL) if,
, 0, (1.1)
Where = 0
.
() 1
Remark 1.1:Life distribution F belongs to ODL if and onlyif < .
()
The rest of the article is structured as follows. The moment inequality developed in section 2, for ODL class of life
distribution based on these inequalities test statistics for testing 0 is exponential against 1 is ODLand not
exponential. In section 3, a new test statistic based on U-statistic is established and have exceptionally Pitman
asymptotic efficiency of some of well-known alternatives, Monte Carlo null distribution critical points are
simulated for sample sizes = 5 1 30 5 50and the power estimates of this test are also calculated at the
significant level = 0.05 for some common alternatives distribution followed by some numerical example. In
section 4, we dealing with right-censored data and selected critical values are tabulated. Finally, the power
estimates for censor data of this test are tabulated and numerical example are calculated.
1
+2 , 0. (2.1)
( + 3) +3
Where
() = = 1 .
0
Proof: Since Fis said to be overall decreasing life ODL class, then from (1.1) we multiplying both sides by
, 0 and integrating over (0, )with respect tot,we get
, (2.2)
0 0
Now,
. . = ,
0
1
= +1 ,
+1 0
1
= +1 ,
+1 0
1
= . (2.3)
+ 1 + 2 +2
Also,
. . = ,
0
1
= +2 ,
+1 +2 0
1
= . (2.4)
+ 1 + 2 + 3 +3
1
= (+2) (3.1)
+ 3 +3
Note that under 0 = 0,and it is positive under 1 .To make the test scale invariant under 0 we use
()
=
+3
It could be estimated based on a random sample 1 , 2 , , from F by
=
+3
1 1 1
= +3 +2 +3 3.2
1 +3
=1 =1
1
Where = =1 is the sample mean, and is estimated by
Setting,
1
1 , 2 = 1 2+2 +3
+3 1
Again and have the same limiting distribution. But since is the usual U-statistics
+3
theory,it is asymptotically normal and all we need to evaluate . The following theorem summarized
+3
the large sample properties of or U-statistic.
Proof: Since and have the same limiting distribution, weconcentrate on
+3
. Now this is asymptotic normal with mean zero and variance2=1, where
1 = 1 , 2 |1 + 2 , 1 |1 (3.5)
But
1
1 = (+2) 1 +3 + 1+2 +2 (3.6)
+3 1
Hence (3.3) follows.Under 0
1
1 = + 2 ! 1 +3 + 1+2 + 2 ! (3.7)
+3 1
1
= 2 3 (3.8)
3
Note that 0 (the exponential distribution) is attained at 0 = 0 in (i), (ii) and at 0 = 1 in (iii), (iv)
Direct calculations for the families in (i) and (ii) are given at = 0.
Using mathematica program to calculate PAE for Weibull family and Gamma family at = 0.
Direct calculations of the asymptotic efficiencies of ODL test are given in Table 1. At = 0.
It is clear from Table 1,we can see that the new ODL test statistic is more efficiency.
www.ijmsi.org 65 | Page
A Moment Inequality For Overall Decreasing Life Class Of Life Distributions With Hypotheses
In view of Table 2, and Fig. 1, it is noticed that the critical values are increasing as the confidence level increasing
and is almost decreasing as the sample size increasing.
www.ijmsi.org 66 | Page
A Moment Inequality For Overall Decreasing Life Class Of Life Distributions With Hypotheses
It is clear from the Table 3 that our test has good powers for all alternatives and the power increases as the sample
size increases. The power is getting as smaller as the ODL approaches the exponential distribution.
Example 1: The data set of 40 patients suffering from blood cancer (Leukemia) from one of ministry of health
hospitals in Saudi Arabia sees Abouammoh et al. [2]. The ordered life times (in years)
0.315 0.496 0.616 1.145 1.208 1.263 1.414 2.025 2.036 2.162
2.211 2.370 2.532 2.693 2.805 2.910 2.912 3.192 3.263 3.348
3.348 3.427 3.499 3.534 3.767 3.751 3.858 3.986 4.049 4.244
4.323 4.381 4.392 4.397 4.647 4.753 4.929 4.973 5.074 4.381
It was found that = 0.678917 and this value exceeds the tabulated critical value in Table 2. It is evident that
at the significant level %95 this data set has ODL property.
Example 2: The following data in keating et al.[21] set on the time, in operating days, between successive failures
of air conditioning equipment in an aircraft. These data are recorded
3.750 0.417 2.500 7.750 2.542 2.042 0.583 1.000 2.333 0.833
3.292 3.500 1.833 2.458 1.208 4.917 1.042 6.500 12.917 3.167
1.083 1.833 0.958 2.583 5.417 8.667 2.917 4.208 8.667
by using (3.3) it is found that = 0.373289 which is less than the critical value of Table 2, then we accept the
null hypothesis. This means that the data set has the exponential property.
Example 3:Using the data set given in Grubbs [19], this data set gives the times between arrivals of 25 customers
at a facility
1.80 2.89 2.93 3.03 3.15 3.43 3.48 3.57 3.85 3.92
3.98 4.06 4.11 4.13 4.16 4.23 4.34 4.73 4.53 4.62
4.65 4.84 4.91 4.99 5.17
It is easily to show that = 0.668617 which is greater than the critical value ofTable 2. Thenwe accept 1
which states that the data set have ODL property and not exponential.
Example 4: Consider the well-known Darwin data (Fisher [18]) that represent the differences in heights between
www.ijmsi.org 67 | Page
A Moment Inequality For Overall Decreasing Life Class Of Life Distributions With Hypotheses
cross- and self-fertilized plants of the same pair grown together in one pot
We can see that the value of test statistic for the data set by formula (3.10) is given by = 2.41461 and this
value greater than the tabulated critical value in Table 2. This means that the set of data have ODLproperty and not
exponential.
1 = .
=
0 = .
Let 0 = 0 < 1 < 2 < < denote the ordered , and is the corresponding to
respectively.
Using the censored data , , = 1, , Kaplan and Meier [20] proposed the product limit estimator.
= 1 = , 0,
+1
:
Now, for testing 0 = 0, against 1 > 0,using the randomly right censored data, we propose the
following test statistic:
1 1
= 3 2 3 .
3
For computational purposes, may be rewritten as
1 1
= . (4.1)
3 3
Where
1 1
= 1 , =2 1 ,
=1 =1 =1 =1
and,
1
2
=3 1 .
=1 =1
Where = .
+1
Table 4.gives the critical values percentiles of test for sample sizes = 5 5 30 10 70,81,86.based on
5000 replications.
www.ijmsi.org 68 | Page
A Moment Inequality For Overall Decreasing Life Class Of Life Distributions With Hypotheses
We noticed that from Table 4 and Fig. 2, the critical values are increasing as the confidence level increasing and is
almost decreasing as the sample size increasing.
We notice from Table 5. We can show that our test has a good power, and the power increases as the sample size
increases.
www.ijmsi.org 69 | Page
A Moment Inequality For Overall Decreasing Life Class Of Life Distributions With Hypotheses
13 14 19 19 20 21 23 23 25 26
26 27 27 31 32 34 34 37 38 38
40 46 50 53 54 57 58 59 60 65
65 66 70 85 90 98 102 103 110 118
124 130 136 138 141 234
The ordered censored observations are:
16 21 44 50 55 67 73 76 80 81
86 93 100 108 114 120 124 125 129 130
132 134 140 147 148 151 152 152 158 181
190 193 194 213 215
Now, taking into account the whole set of survival data (both censored and uncensored). It was found that the
value of test statistic for the data set by formula (4.1) is given by = 0.0492696 and this value greater than the
tabulated critical value in Table 5. This means that the data set hasODL property.
Example 2.
On the basis of right censored data for lung cancer patients from Pena [29]. These data consists of 86 survival
times (in month) with 22 right censored:
The whole life times (non-censored data):
0.99 1.28 1.77 1.97 2.17 2.63 2.66 2.76 2.79 2.86
2.99 3.06 3.15 3.45 3.71 3.75 3.81 4.11 4.27 4.34
4.4 4.63 4.73 4.93 4.93 5.03 5.16 5.17 5.49 5.68
5.72 5.85 5.98 8.15 8.62 8.48 8.61 9.46 9.53 10.05
10.15 10.94 10.94 11.24 11.63 12.26 12.65 12.78 13.18 13.47
13.96 14.88 15.05 15.31 16.13 16.46 17.45 17.61 18.2 18.37
19.06 20.7 22.54 23.36
The ordered censored observations are:
11.04 13.53 14.23 14.65 14.91 15.47 16.49 17.05 17.28 17.88
17.97 18.83 19.55 19.58 19.75 19.78 19.95 20.04 20.24 20.73
21.55 21.98
We now account the whole set of survival data (both censored and uncensored), and computing the test statistic
given by formula (4.1). It was found that = 0.433052 which is exceeds the tabulated value in Table 5. It is
evident that at the significant level 0.95 .Then this data set has ODL property.
REFERENCES
[1]. Abdul Alim, A.N and Mahmoud, M.A.W. A goodness of fit approach to for testing NBUFR (NWUFR) and NBAFR (NWAFR)
properties. InternationalJournal of Reliability and Applications (2008); 9:125--140.
[2]. Abouammoh,A. M., Abdulghani, S. A. and Qamber, I. S. (1994). On partial orderings and testing of new better than used classes,
Reliability Eng. Syst. Safety, 43, 37-41.
[3]. Abu-Youssef, S. E. (2002). A moment inequality for decreasing (increasing) mean residual life distributions with hypothesis testing
application, Statist. Probab.Lett., 57, 171-177.
[4]. Ahmad,I.A. (2001). Moments inequalities of ageing families of distribution with hypothesis testing applications, J. Statist. Plan. Inf.,
92,121-132.
[5]. Ahmed, A. N. andAbouammoh, A. M. The new better than used class of life distribution, Advances in Applied Probability, 20,
237-240 (1988).
[6]. Ahmad, I.A. (1994). A class of statistics useful in testing increasing failure rate average and new better than used life distribution.
J.Statist. Plant. Inf., 41, 141-149.
[7]. Ahmed,A.N and Abouammoh,A.M. (1992). On renewal failure rate classes of life distributions. Statist.AndProbab.Lett. , 14, 211-187.
[8]. Aly,E.E.I(1989).On testing exponentiality against IFRA alternative, Metrika, 36, 225-267.
[9]. Barlow, R.E., Marshall, A.W.andProschan, R.(1963). Properties of probability distributions with monotone hazard rate. Ann. Math.
Statist., 34, 375-389.
[10]. Barlow,R.E.andProschan, F. (1981). Statistical Theory of Reliability and Life Testing.To Begin with Silver Spring, M D.
[11]. Deshpande, J. V., Kochar, S.C. and Singh, H., Aspects of positive aging, Journal of Applied Probability, 288, 773-779 (1986).
www.ijmsi.org 70 | Page
A Moment Inequality For Overall Decreasing Life Class Of Life Distributions With Hypotheses
[12]. Diab,L.S.etal(2009). Moments inequalities for NBUL distributions with hypotheses testing applications, Contem. Engin. Sci., 2,
319-332.
[13]. Diab,L.S. (2010). Testing for NBUL using goodness of fit approach with application. Stat. papers., 51, 27-40.
[14]. Diab,L.S.(2013). A new approach to moments inequalities for NRBU and RNBU classes with hypothesis testing applications.
International Journal of Basic &appliedSciences IJBAS-IJENS .Vol. 13, No.06.pp.7-13.
[15]. Diab,L.S. and Mahmoud, M.A.W.(2007). On testing exponentiality against HNRBUE based on a goodness of fit. International
Journal of Reliability and Applications.8, No.1,27-39.
[16]. EL-Arishy,S.M., Mahmoud, M. A. W. and Diab L. S. (2003). Moment inequalities for testing new renewal better than used and
renewal new better than used classes, Int. J. Rel. Appl., 4, 97-123.
[17]. EL-Arishy,S.M., Mahmoud, M. A. W. and Diab L. S. (2005). Testing renewal new better than used life distributions based on u-test,
Appl. Math. Model., 29, 784-796.
[18]. Fisher,R.A. (1966). The Design of Experiments, Eight edition, Oliver & Boyd, Edinburgh.
[19]. Grubbs, F. E. (1971). Fiducial bounds on reliability for the two parameter negative exponential distribution. Technomet., 13, 873-876.
[20]. Kaplan,E.L.and Meier, P.(1958). Nonparametric estimation from incomplete observation. J. Amer. Statist. Assoc., 53,457-481.
[21]. Keating, J.P., Glaser, R.E. and Ketchum, N.S. (1990).Testing hypothesies about the shape of a gamma distribution Technometric, 32,
67-82.
[22]. Klefsjoe, B. (1981). HNBUE survival under some shock models, Scand J. Statist, 8,39-47.
[23]. Klefsjoe, B. (1982). The HNBUE and HNWUE classes of life distributions.Naval, Res. Log.Quart., 29, 331-344.
[24]. Koul,H. L. (1977 a). A test for new better than used. Comm. Statist. A-Theory Methods, 6, 563-574.
[25]. Kumazawa, Y. (1983). Testing for new better than used. Comm. Statis. Theor. Meth.,12, 311-321.
[26]. Mahnoud, M. A. W. and Diab,L. S. (2008). A goodness of t approach to decreasing variance resudual life class of life distributions.
JSTA, 7, No. 1, 119-136.
[27]. Mahmoud, M. A. W., Albasam, M. S. and Abdulfattah, E. H. (2010).A new approach to moment inequalities for NBRU class of life
distributions with hypothesies testing applications. International journal of reliability and applications.V.11, No 2, pp.139-151.
[28]. Proschan,F.andPyke, R. (1967). Tests for monotone failure rate. Proc. 5 th Berkeley Symp. Math Statist.Probab., III, 293-312.
[29]. Pena, A. E. (2002).Goodness of fit tests with censored data.http://statman Stat.sc.edupenajtajkspresentedjtalkactronel.
[30]. Sepehrifar, M., Yarahmadiany, S. andYamadaz, R. (2012). On classes of life distributions: Dichotomous Markov Noise shock Model
with hypothesis testing applications. Math. ST., arXiv:1210.0291v1.
[31]. Siddiqui,M.M. and Bryson, M. C .(1969). Some criteria of ageing . J. Amer. Statist. Assoc., 64, 1472-1483.
[32]. Susarla, V. and Vanryzin, J (1978). Empirical bayes estimations of a survival function right censored observation. Ann. Statist., 6,
710-755.
[33]. Zacks,S. (1992). Introduction to Reliability Analysis Probability Models and Methods. Springer Verlag, New York.
www.ijmsi.org 71 | Page